With over 4 billion internet users worldwide, engaging in online videos and entertainment remains one of the most popular activities among users which led to a high need for video optimization over modern cellular network .
In this article, we are describing some of the key factors which can help to ensure smooth video streaming (lower access time and lower stalling should be the key factor to ensure a good quality of experience on a modern network.
RF condition
Starting by the most common factor for poor video performance is poor RF condition, as it leads to lower MCS allocation or high Retransmission while modern video applications rely on the ABR features to adapt to different RF conditions and available bandwidth change along with initial resolution discovery to set initial playing resolution, sudden RF fluctuating (on FR2 for example ) may not be caught by those features due to higher window for evaluation.
MAC & Link adaptation
Moving to the high layer, retransmission, and MCS allocation may have a different impact on QOE performance, setting a BLER target on an aggressive target, for example, provides a too-conservative MCS decreasing the user throughput and available bandwidth without a linear gain on the converged BLER value however too relaxed target increases the retransmission excessively and may have a negative impact depends on adopted TCP congestion approach even if NR leg relies on Flexible Timing between Data Transmission and HARQ response it is not the case on LTE leg which still rely on fixed retransmission timing.
Split bearer flow control performance
There are two factors to consider when coming to split bearer performance during flow control, lower latency ( lower delay for packet delivery to higher layer ) and maximum bandwidth when ul-DataSplitThreshold specified in TS 36.323 represent the threshold value for uplink data split, downlink split bearer is relying on The DL Data Delivery Status specified in TS 36.425 which is used by the gNB to get the performance over X2-U for the packet ( expected throughput, highest PDCP SN submitted ..etc. )
several algorithms are used by different vendors for downlink delay estimation on split bearer implementation however a proper delay threshold value should be tunned which can maximize bandwidth usage and limit delay to higher layers while switching between LTE and NR legs, especially while moving toward 5G coverage cell edge ensuring optimized packet distribution and maximizing performance
TCP Congestion control algorithm.
Starting from the higher layer on the transport layer TCP is one of the main standards that define the internet Many modern networks which relies on a modern implementation, even if Google implemented QUIC ( on the TOP of UDP ) which solves a number of transport-layer and application-layer problems experienced by modern web applications it remains very similar to TCP and still many Video application seems to be relying on TCP for the initial session establishment while starting connection on TCP and advertising for QUIC seems to be still more reliable approach for many applications, there is a lot to discuss TCP so a separate future topic will be dedicated to it
TCP uses a congestion window and a congestion algorithm to avoid congestion which falls under different categories: Delay based & loss-based & Combined loss and delay based.
Delay-based Algorithms will tend to provide lower delay reducing the probability of video stalling by limiting the standing queue but under certain circumstances (high packet loss for example) provide lower bandwidth utilization and low effective throughput whereas loss-based will work on complete filling the queue maximizing the bandwidth utilization but creating higher standing queues which may be lead to stalling and react negatively in case of high packet loss caused by non-congested network
In the case of network-implemented TCP optimization relying on Cubic or BBR remains the most reliable protocol to be adopted depending on the network broadband, cashing, and packet loss level, BBR will ensure smooth video playing due to its reliability on BDP, and cubic in case of high packet loss network experiencing several packet losses due to congestions.
Commentaires